Posture Prediction and Optimization - Human Motion Simulation: Predictive Dynamics - page 53

2 2

where ð m i g Þ represents the weights and ð Δ h i Þ acts as the individual objective

functions; is the number of lumped masses. Potential energy is a relative quan-

tity, and the term

κ

h i usually refers to the vertical distance between a mass and a
plane of zero potential, where the plane is the same for all masses. However, in
this case,

Δ

h i is measured relative to the neutral position for a particular mass.
Essentially, each mass has a different plane of zero potential. In this case, the ini-
tial position is the neutral position described in relation to joint displacement.
Thus, with this performance measure, the human model again gravitates towards
the neutral position, as was the case with joint displacement. However, horizontal
motion of the lumped masses has no effect.

Δ

3.7.4 Discomfort
The idea of modeling discomfort can be somewhat ambiguous, as it is a subjec-
tive quantity, the evaluation of which may vary from person to person. However,
it is possible to incorporate distinct factors that contribute to discomfort, though
the actual absolute value for discomfort may not be significant. In this section,
we present a human performance measure for musculoskeletal discomfort that
incorporates three such factors:

1. The tendency to move different segments of the body sequentially
2. The tendency to gravitate to a reasonably comfortable neutral position
3. The discomfort associated with moving while joints are near their respective
limits.

Some experimental work has been completed on discomfort in an effort to
determine which factors contribute to discomfort, and the above-mentioned fac-
tors surface in these studies. However, these factors have not yet been collectively
incorporated into an effective optimization-based human performance measure for
use with virtual humans.
In order to incorporate the first factor (the tendency to move different seg-
ments of the body sequentially), there are several strategies to induce motion in a
certain order, or with higher weighted joints than others.
Forexample,inanefforttoreachaparticular target point, one first uses one's
arm. Then, only if necessary, does one bend the torso. If the target is still out of
reach, one may extend the clavicle joint or take a step to reach it. The
weights used to approximate the lexicographic approach are shown in Table 3.3
( Marler et al., 2005a,b , 2009).
The weights in Table 3.3 are used in a function that is based on Equation (3.7)
with the neutral position defined as shown in Equation (3.8) . In this way, the sec-
ond factor of discomfort (the tendency to gravitate to a reasonably
comfortable neutral position) is incorporated. Prior to applying the weights, each
term in Equation (3.7) is normalized as follows:

q i

q i 2

q norm Δ 5 (3.12)

q i 2 q i